Elsevier

Social Science & Medicine

Volume 60, Issue 2, January 2005, Pages 409-419
Social Science & Medicine

The causes of low back pain: a network analysis

https://doi.org/10.1016/j.socscimed.2004.05.013Get rights and content

Abstract

Beliefs regarding the cause of low back pain differ between individual sufferers and health care professionals. One consequence of this is the potential acquisition of maladaptive attitudes and behaviour in relation to pain, and increases in the utilisation of primary care services (Health Expect.3(3) (2000) 161). Methods that have been used to elicit the causal interpretation of social phenomena are varied yet they are unable to categorically demonstrate the different weightings or levels of importance that individuals may assign. The diagram method of network analysis allows individuals to spontaneously consider pathways they believe to be critical to a target event and to determine the strength of those pathways.

Seventy-one completed diagrams indicating the causes that sufferers perceived to be related to low back pain were analysed. The mean number of direct causal paths was 5.61 (SD=3.25) and mean number of indirect causal links was 1.16 (SD=2.34). A significant correlation between path frequency and path strength was also found (r=0.76, p=0.001). Sufferers do not have an overtly complex view of the causative factors of low back pain but were able to define four core contributory causes (disc, sciatica, lifting, and injury) and one indirect pathway between lifting and injury. There was a clear delineation between external (biomedical) and internal (person-related) factors that were attributed to low back pain acquisition. By determining these causal attributions it is proposed that treatment packages could be tailored to address biases in thinking. This may be particularly useful for those individuals who attribute their pain as a consequence of external (or biomedical) causes.

Introduction

Low back pain is a pervasive health care problem. In 1993, the prevalence of back pain involved sixteen and a half million people in the United Kingdom; the annual costs of back pain to the National Health Service were £480 million, whilst the costs of Department of Social Security benefits were estimated at £1400 million (Clinical Standards Advisory Group, 1994). Since these data were reported further economic analysis has revealed that the incidence is rising, and with it the associated costs. Maniadakis and Gray (2000) proposed that in 1998, in the United Kingdom, the direct health care cost of back pain was £1632 million and the cost of informal care and related production losses totalled £10668 million. The prevalence appears to be increasing and this may be associated with cultural changes that have influenced people's awareness of more minor back symptoms and their willingness to report them (Palmer, Walsh, Bendall, Cooper, & Coggan, 2000) and the attitudes and beliefs held by referring physicians (Buchbinder, Jolley & Wyatt, 2001).

Chronic low back pain is defined as pain within the lumbosacral region, buttocks and thighs that is ‘mechanical’ in nature: it varies with physical activity and varies with time (Waddell, 1998). Additionally, it is seen to be pain that persists beyond twelve weeks (Clinical Standards Advisory Group, 1994) or that lasts beyond the expected period of healing (Andersson, 1999). However, as highlighted by Evans and Richards (1996) there is no standard or agreed definition of chronic low back pain and there remains continuing uncertainty and disagreement within the medical professions related to its cause, nature and treatment. This is problematic, not least because a failure to reach consensus regarding the cause and nature of it can lead to inconsistencies in the advice and subsequent treatment that is given to individuals suffering with low back pain. Such inconsistencies may also shape the beliefs held by the sufferers in relation to their pain. Furthermore, aside from the physical sensation of pain, chronic low back pain is characterised by the impact that it has on the individual's life. Associated anxiety, depression and stress are frequently cited as having an additional major impact on the lives of those who live with persistent low back pain (Croft et al., 1995; Main, 1983; Sullivan, Reesor, Mikail, & Fisher, 1992; Averill, Novy, Nelson, & Berry, 1996). The presence of pain is seen as a symptom that requires the help of a physician and may lead to the search for a cure as long as the pain continues (McCracken, 1998). Patients who are more obdurate in their attempts to avoid or rid themselves of pain, and consequently view pain as an absolute barrier to a better life suffer more ill consequences. Those individuals who are unaccepting of their pain appear to be facing the distress that comes from attempting to control an unchangeable aversive experience (Thompson, 1981).

Despite this emerging body of predominantly biomedical evidence that attempts to explain or account for the continuance of pain, an equally valid perspective is proposed that challenges the ‘dominant medico-scientific’ approach (Williams & Bendelow, 1998, p. 169) to the study of pain. This different point of view questions the validity of the dominant dualist explanations provided to construe pain, whereby ‘pain and meaning’ (Morris, 1991, p. 44) are viewed separately, when in fact they should be seen as interdependent. Moreover, Morris (1991) suggested that when we begin to explore the multiple meanings that can be assigned to the pain experience these often prove to be ‘at least as important as the medicines we consume’ (p.44). There is a clear need therefore, for the meanings that sufferers ascribe for their pain and beliefs relating to its cause to be explored.

Beliefs are assumptions about reality that serve as a perceptual lens or a ‘set’ through which events are interpreted (Lazarus & Folkman, 1984). In relation to health and illness evidence attests to the differing beliefs that are held between the lay populace and the medical profession to account for the causes of wellness and disease (Herzlich, 1973; Pill & Stott, 1982; Blaxter, 1982; Helman, 1984; Davison, Davey Smith, & Frankel, 1991). Early explanations to account for these differences centred on the social psychological models of help-seeking behaviour and decision-making (Uehara, 2001) and the prominence of the mind-body dualism entrenched within the traditional biomedical culture (see Ferguson & Marras, 1997). However, Williams (1984) suggests that lay theories of aetiology serve to ‘reconstruct some sense of order from the fragmentation produced by illness’ (p. 177) and in so doing common sense explanations of illness can be seen to develop to provide a vehicle to enable this reconstruction to occur. Peters, Stanley, Rose, and Salmon (1998) nevertheless suggest that lay and medical beliefs are not separate systems and echo Shorter's (1992) view that medical accounts of common physical symptoms continually diffuse into lay discourse about biology and illness. Low back pain is one such common physical complaint where this diffusion is evident. However, it must be noted that there are conceptual differences between lay knowledge (where lay people may have particular insights into the phenomenon of low back pain) and the phenomenon specific knowledge that patients or sufferers of low back pain have as a direct consequence of their pain experience. Whilst both are equally valid it is the latter that places patients and sufferers in a unique position because of this insider knowledge (Arskey, 1994).

Beliefs regarding the cause of low back pain differ between individual sufferers and health care professionals (Cedraschi, Reust, Roux, & Vischer, 1992; Skelton, Murphy, Murphy, & O’Dowd, 1995; Hermoni et al., 2000). Borkan, Reis, Hermoni, and Biderman (1995) found that individual sufferers articulated several factors to account for their pain and these formed the basis for the beliefs that they held. The diversity of these factors ranged from explanations that the presence of pain was due to: an injury; degeneration of the spine; as a punishment; stress and the weather. Additionally, factors such as ‘work’ ‘ageing’ and ‘childhood injuries’ (p.983) were seen to constitute the primary causes of low back pain. More recent findings suggest that perceptions about the cause of low back pain being due to ‘a slipped disc’ or ‘a trapped nerve’ are factually incorrect and may account for increased utilisation of primary care services (Moffett, Newbronner, Waddell, Croucher, & Spear, 2000). This demonstrates the point highlighted by Shorter (1992) whereby medical terminology becomes ingrained into everyday lay language; one consequence of which is that a poor understanding of the pain symptoms may lead to the acquisition of maladaptive attitudes and behaviour in relation to the pain (Geisser & Roth, 1998).

Methods that have been used to elicit the causal interpretation of social phenomena are varied and include semi-structured interviews (Antaki, 1985; Litton & Potter, 1985) and spontaneous discourse (Campbell & Muncer, 1987). Explication of beliefs and meaning regarding illness perception has previously occurred via an exploration of lived experience. Utilising patient narrative accounts Kleinman (1988) explored and mapped an explanatory lay model to elucidate how illness conditions were conceptualised and how these conceptualisations impacted on treatment choice, compliance and outcome. Other research has shown how individuals perceive hypertension (Greenfield, Borkan, & Yodfat, 1987) comprehend hip fracture (Borkan, Quirk, & Sullivan, 1991) and make sense of everyday pain (Aldrich & Eccleston, 2000). Through direct questioning Underwood (1997) was able to determine patients’ beliefs as to the cause of their low back pain. Whilst research evidence provides insight to the extent of lay knowledge, Furnham (1994) argues that the structure and determinants of lay beliefs about health and illness are not considered. He contends that scrutiny is needed to understand those factors that are pivotal to the construction of the meanings given. In support of this position Aldrich and Eccleston (2000) highlight the paucity of detailed work that describes how meaning is created. One mechanism by which this can be achieved is through the exploration of causal attributions. Roesch and Weiner (2001) stated that attributions were constructs worthy of consideration because they allowed an understanding of the world as it had occurred in the past (both recent and distant) but they also served to guide future behaviour. One caveat to this approach is offered by Peterson and Seligman (1984) and Forsterling (1986) whom have highlighted the role that personal biases may play in the (mis)interpretation of events. Russell (1998) however, has argued that the most fair and ethical approach to a consideration of health outcome is to use the perceptions of the service users as it places delivered care in the context of people's expectations. It is within this context that condition-specific attributions should be explored.

Whilst useful, these methods of inquiry are unable to categorically demonstrate different weightings or levels of importance that individuals may assign to any given set of condition-specific causative determinants. Consequently certain factors may be deemed as less important or irrelevant and this may not be elicited through the analyses chosen. More recently however, network analysis has been utilised to overcome some of these issues.

Early network studies employed a grid method (see Fig. 1) to determine whether respondents perceived there to be a link between causal factors in relation to one target variable (Litton & Potter (1985), Heaven (1988); Heaven, 1994).

Participants are presented with a grid in which the proposed causes of a social phenomenon (such as employment, loneliness or poverty) are presented vertically and horizontally along the top and sides of a grid. Individual respondents are asked to determine whether there is a link between each of the factors. Agreement was originally indicated by either a binary scale in which individuals enter a ‘1’ if they perceive a link or a ‘0’ if they do not (see Lunt, 1988; Campbell & Muncer, 1990; Lunt & Livingstone, 1991; Muncer, Epro, Sidorowicz, & Campbell, 1992).

More recent studies have tended to replace the binary form with a Likert scale in which participants rated the strength of a link on a scale of 1–5 (see Lunt, 1991; Heaven, 1994). Individual grids are then aggregated and networks constructed hierarchically with the strongest link entered first. Network construction stops according to a cause-to-link ratio calculation in which new links are added to the network provided that this does not entail adding a large number of other causal links from causes that already appear on the network.

Criticisms of this method centre on the fact that the resulting networks are not the production of any one individual but instead represent an aggregated response format (Muncer & Gillen, 1997). They are often, therefore, not representative of what participants perceive as the causal links and generally produce larger networks with more causal links than are perceived by any one participant (Muncer & Gillen, 1997). Additionally, the grid method requires respondents to consider possible links between causes that may account for the target variable (e.g., loneliness) rather than direct links between the causes and the target variable. It has subsequently been found on many occasions that the direct links between the causes and the target are often considered more important than the links between causes (Heffernan, Green, McManus, & Muncer, 1998; Muncer, Taylor, & Ling, 2001b). Furthermore, a target factor may have reciprocal causal links with the nominated causes. In a recent network study of the causes of health and illness, participants’ representations suggested that stress was an important cause of illness and that illness was an important cause of stress (Muncer et al., 2001b). Preventing individuals from representing direct paths (as in the earlier grid method) may yield an over complex representation on the one hand (see, for instance, the representation elicited in Lunt, 1991), as individuals seek to encode their perceptions within the constraints of proposed causes, and, on the other hand, it may limit the subtlety of the representation by precluding the representation of connections to the target factor that are both direct and indirect from a given causal factor. Likewise, there is no place within the grid method for free choice by respondents to include factors that they may perceive as relevant but are not included within the framework of the grid. The diagram method (see Fig. 2) proposed by Green and McManus (1995) circumvents these issues.

Green and McManus (1995) required individuals to draw a network diagram of the risk factors for a target factor, coronary heart disease. Unlike previous network elicitation studies, individuals were not forced to consider each and every connection amongst the factors but were free to sample as they wished. Individuals represented a causal relationship between two factors by drawing a line connecting them and indicated the direction of the causal influence using an arrowhead. Also in contrast to previous studies, the target factor (coronary heart disease) was explicitly represented. In the diagrams created, a possible causal factor may be connected to the target factor in a variety of ways. A causal factor may have a direct path to the target factor, or it may have an indirect path to it via some other factor, or it may have both a direct path and an indirect path to the target factor. The diagram represents what individuals spontaneously consider the critical pathways.

In addition to representing a path, individuals were required to rate the strength of each causal path on a scale from zero to one hundred. A composite diagram was constructed that indicated both the percentage of individuals including each causal path and the mean strength of those causal paths. Green and McManus (1995) showed that the total path strength of a factor (the strength of both direct paths and all indirect paths) predicted participants’ ratings of the effectiveness of different actions based on each of the risk factors in reducing the risk of coronary heart disease.

In a subsequent study that examined perceptions of the factors increasing a person's prospects of employment, Green, McManus, and Derrick (1998) confirmed the importance of path strengths in predicting the ratings of the effectiveness of different actions designed to increase a person's employment prospects. Muncer, Taylor, Green, and McManus (2001a) have also used the diagram drawing method to investigate nurses’ representations of the causes of work-related stress. In this study they also investigated the reliability and validity of the resulting networks by comparing the network produced with that produced from a previous network study and also with the findings of a large-scale study of stress in the National Health Service (Haynes, Wall, Boplden, Stride, & Rick, 1999). In the latter case, the pattern of correlation between the nine dimensions of the causes matched many of the features of the network diagram.

Within the diagram method of network construction individuals are presented with a list containing factors that may cause the target variable but additionally there is opportunity to incorporate other factors that the respondent may feel is personally relevant. Individual respondents are required to draw a line to represent the link between the causal factor and the target variable. They are also required to indicate the strength of this cause by providing a value from, for example, 0 to 100 (Green et al., 1998). Of particular relevance to these network studies is that to date, the majority have utilised samples of participants that are comprised of undergraduate students. As a result, these studies, whilst demonstrating the efficacy of the methodology, may be problematic due to the elicitation of causal attributions to social phenomena from individuals who do not perceive the target variable to be personally relevant at the time of grid construction (e.g., loneliness). The exception to this are the studies which explored factors that were perceived to be contributory to work-based stress in nurses and utilised a sample consistent with the research aim (Taylor, White, & Muncer, 1999; Muncer, Epro, Sidorowicz, & Campbell (1992), Muncer, Taylor, Green, and McManus (2001a)). It could be argued therefore, that personally experienced phenomena would hold a different meaning in comparison to those who are asked to consider an abstract or personally meaningless concept and consequently this may impact on the network that is derived.

No studies to date have utilised network analysis to elicit those factors that low back pain sufferers perceive cause low back pain. This research is important for two reasons. Firstly, it is needed to develop our understanding of the complexities inherent in the casual attributions sufferers make to account for their pain. Secondly, it may provide additional evidence to support claims made by Moffet et al. (2000) that perceived causes that are factually incorrect may lead to greater utilisation of health care services.

Section snippets

Participants

Two hundred and thirty four low back pain sufferers comprised the initial sample pool. These patients were recruited as part of a larger study exploring expectations of low back pain treatment. Consecutive patients referred to two North East of England Spinal Assessment Clinics and two North East of England Pain Clinics comprised the study participants. One element of this larger study required individuals to indicate what they believed the cause of their back pain to be. One hundred and eleven

Results

A composite diagram was prepared by examining the separate paths between any two factors for each of the 71 participants. Participants produced mainly direct paths between the causes and the target back pain, with only 35.8% of participants including indirect causal paths between causes. The mean number of direct causal paths was 5.61 with a standard deviation of 3.25 and the mean number of indirect causal links was 1.16 with a standard deviation of 2.34. Causal paths that were endorsed by at

Content analysis

In addition to the elicitation of the main causative pathways, quantitative content analysis was performed on the other causes that individual sufferers perceived also contributed to the cause of their low back pain. Twenty-seven participants provided a total of 33 responses that account for 16 other causes of low back pain. These are shown in Table 3.

Similar to the main causes that were extracted via the network analysis, the related perceived causes of low back pain were attributed to either

Discussion

The findings from this study have shown that sufferers did not have an overtly complex view of the causative factors of low back pain but were able to define four core (high consensus ⩾50% consensus) contributory causes of low back pain (disc, sciatica, lifting, and injury) and one indirect pathway between lifting and injury. Moreover, the total number of pathways that were generated (n=16) reflects the inherent difficulties found in clinical practice in providing a concise diagnosis for pain

Conclusion

This network analysis has revealed the complexities, subtleties and consensus of the causal attributions made by low back pain sufferers. This is the first study, to our knowledge, to utilise this method with individuals who have chronic low back pain and it has shown that the individuals within our sample have a good understanding of their pain symptoms and are able to attribute these to simple diagnostic criterion. Utilisation of the diagram method with a sample that has personal experience

References (72)

  • L.M. McCracken

    Learning to live with the painAcceptance of pain predicts adjustment in persons with chronic pain

    Pain

    (1998)
  • C.J. Main

    The modified somatic perception questionnaire

    Journal of Psychosomatic Research

    (1983)
  • N. Maniadakis et al.

    The economic burden of back pain in the UK

    Pain

    (2000)
  • C.R. May et al.

    Dealing with doubtHow patients account for non-specific chronic low back pain

    Journal of Psychosomatic Research

    (2000)
  • S. Peters et al.

    Patients with medically unexplained symptomsSources of patients’ authority and implications for demands on medical care

    Social Science and Medicine

    (1998)
  • W.G. Pickering

    Does medical treatment mean patient benefit?

    Lancet

    (1996)
  • R. Pill et al.

    Concepts of illness causation and responsibilitySome preliminary data from a sample of working class mothers

    Social Science and Medicine

    (1982)
  • S.C. Roesch et al.

    A meta-analytic review of coping with illness—Do causal attributions matter?

    Journal of Psychosomatic Research

    (2001)
  • E. Russell

    The ethics of attributionThe case of health care outcome indicators

    Social Science and Medicine

    (1998)
  • M.J.L. Sullivan et al.

    The treatment of depression in chronic low painReview and recommendations

    Pain

    (1992)
  • E. Uehara

    Understanding the dynamics of illness and help-seekingEvent-structure analysis and a cambodian-american narrative of ‘spirit invasion’

    Social Science and Medicine

    (2001)
  • C. Antaki

    Ordinary explanations in conversationCausal structures and their defence

    European Journal of Social Psychology

    (1985)
  • H. Arskey

    Expert and lay participation in the construction of medical knowledge

    Sociology of Health and Illness

    (1994)
  • M. Blaxter

    Mothers and daughtersA three-generational study of health attitudes and behaviour

    (1982)
  • R. Buchbinder et al.

    Population based intervention to change back pain beliefs and disabilityThree part evaluation

    British Medical Journal

    (2001)
  • A. Campbell et al.

    Models of anger and aggression in the social talk of women and men

    Journal for the Theory of Social Behaviour

    (1987)
  • C. Cedraschi et al.

    The role of prior knowledge on back-pain education

    Journal of Spinal Disorders

    (1992)
  • Clinical Standards Advisory Group (1994) Report on Back Pain. London:...
  • P.R. Croft et al.

    Psychologic distress and low back painEvidence from a prospective study in the general population

    Spine

    (1995)
  • C. Davison et al.

    Lay epidemiology and the prevention paradoxThe implications of coronary candidacy for health education

    Sociology of Health and Illness

    (1991)
  • Evans, G.,Richards, S. (Eds), (1996). Low Back Pain: An evaluation of therapeutic interventions. Health Care Evaluation...
  • F. Forsterling

    Attributional conceptions in Clinical Psychology

    American Psychologist

    (1986)
  • D.P. French et al.

    The structure of beliefs about the causes of heart attacksA network analysis

    British Journal of Health Psychology

    (2002)
  • M.E. Geisser et al.

    Knowledge of and agreement with chronic pain diagnosisRelation to affective distress, pain beliefs and coping, pain intensity and disability

    Journal of Occupational Rehabilitation

    (1998)
  • D.W. Green et al.

    Cognitive structural modelsThe perception of risk and prevention in coronary heart disease

    British Journal of Psychology

    (1995)
  • D.W. Green et al.

    Cognitive structural models of unemployment and employment

    British Journal of Social Psychology

    (1998)
  • Cited by (44)

    • Low back pain

      2021, The Lancet
      Citation Excerpt :

      Such associations not only incite pain-related fear, they also can lead to hyperalgesia.57 Erroneous beliefs about the relationship between particular movements and pain are prevalent in patients with low back pain,58 but are also found among health professionals.59 For example, the use of expressions implying harm (eg, “Your spine looks like that of a 70-year-old”) could inadvertently evoke pain-related fear.

    • A Study on the Effect of Soft Wearable Suit Using Elastic Band

      2022, Journal of the Korean Society for Precision Engineering
    View all citing articles on Scopus
    View full text